Psychopharmacology

, Volume 227, Issue 3, pp 425–436

Caffeine increases liking and consumption of novel-flavored yogurt

Authors

  • Leah M. Panek
    • Department of Exercise and Nutrition Sciences, School of Public Health and Health ProfessionsUniversity at Buffalo
  • Christine Swoboda
    • Department of Exercise and Nutrition Sciences, School of Public Health and Health ProfessionsUniversity at Buffalo
  • Ashley Bendlin
    • Department of Exercise and Nutrition Sciences, School of Public Health and Health ProfessionsUniversity at Buffalo
    • Department of Exercise and Nutrition Sciences, School of Public Health and Health ProfessionsUniversity at Buffalo
Original Investigation

DOI: 10.1007/s00213-013-2971-6

Cite this article as:
Panek, L.M., Swoboda, C., Bendlin, A. et al. Psychopharmacology (2013) 227: 425. doi:10.1007/s00213-013-2971-6

Abstract

Rationale

Caffeine has been shown to increase preference for beverages with which it is paired; however, it is not known if caffeine alters liking for foods with which it is paired indirectly.

Objectives

The purpose of the current experiment was to test the hypothesis that a caffeinated beverage paired with a novel-flavored yogurt will increase preference for that yogurt compared to one paired with placebo. We also tested the hypothesis that liking would increase more when caffeine was paired with high energy density yogurt.

Methods

Men and women (n = 62) were randomized to receive a beverage containing placebo (PLA) or caffeine (CAF) and to consume a low (LED) or high energy density (HED), novel-flavored yogurt. Participants rated, ranked, and consumed seven novel-flavored yogurts and then had a target yogurt paired with either PLA or CAF over four consecutive days.

Results

In general, yogurt liking increased over time, the HED yogurt was liked more than the LED yogurt, and yogurt paired with caffeine was liked more than yogurt paired with placebo. Participants showed a significant increase in liking of LED yogurt paired with caffeine compared to those with LED yogurt paired with placebo.

Conclusions

Caffeine administration may increase liking and consumption of novel-flavored foods, particularly if the food is not highly liked at baseline. This suggests that caffeine pairing may be a way to increase liking of LED foods, such as vegetables and fruit.

Keywords

CaffeineEnergy densityHedonicsFlavor–nutrient learningConditioning

Introduction

Caffeine use is prevalent among adults, with common sources being coffee, tea, and carbonated beverages. Caffeine has dose-dependent effects on physiology, mood, and subjective responses with moderate doses associated with positive effects and high doses associated with adverse, anxiogenic effects (Keast et al. 2011; Yeomans et al. 2008). While caffeine exists naturally in beverages such as coffee and tea, it is present as an additive in other beverages, such as soda and energy drinks. The purpose of exogenous caffeine in soda and energy drinks is unknown, but many have speculated that it enhances hedonic and reinforcing effects of beverages with which it is paired, which increases consumption and sales of those beverages (Garber and Lustig 2011; Griffith and Vernotica 2009). The pairing of caffeine with beverages requires manufacturers to increase the concentration of sweetener to mask the bitter flavor of the caffeine, which may promote increased energy intake, increased weight, and increased BMI (Keast et al. 2011).

Caffeine positively influences preferences for beverages with which it is paired (Rogers et al. 1995; Smit et al. 2006; Yeomans et al. 2005, 2007, 2008). For example, the addition of caffeine to novel-flavored beverages results in an increased flavor preference for these beverages relative to the same beverages paired with placebo (Chambers et al. 2007; Dreumont-Boudreau et al. 2008; Richardson et al. 1996; Yeomans et al. 2002a, 2005, 2007). Changes in beverage preference after pairing with caffeine may depend on familiarity with the beverage. For example, changes in flavor preference with exposure to caffeine occur more readily in drinks that are unfamiliar (Smit et al. 2006). The increase in hedonic ratings for caffeine-paired beverages has also been shown to relate to reversal of symptoms associated with caffeine withdrawal and is not seen in low/nonusers or in nonwithdrawn consumers (Chambers et al. 2007; Richardson et al. 1996; Rogers et al. 1995). For example, typical afternoon caffeine consumers showed an increased preference for a caffeine-containing drink after lunch compared with nonconsumers (Richardson et al. 1996). It is possible that repeated pairing of caffeine with sweet taste, as what occurs with soda and sweetened coffee drinks, may also result in increased preferences for sweet foods and beverages. For example, adults who regularly consume caffeinated beverages appear to also consume higher amounts of sweet and fat food and lower amounts of healthy foods (Sanchez-Villegas et al. 2009; Yeomans et al. 2007). More research is needed to confirm and replicate these findings and determine the extent to which caffeine can alter preferences for foods and beverages.

Another factor that influences the liking of novel foods and beverages is the flavor and nutrient content of the foods with which they are paired. Flavor–flavor learning occurs at the orosensory level when a neutral or novel flavor is paired with a flavor that is already perceived as pleasant in order to increase liking of a novel flavor (Temple et al. 2009; Yeomans et al. 2008). Flavor–nutrient learning occurs in a fasted state at the gastrointestinal level when the body recognizes an essential nutrient in a consumed food and associates the novel flavor with the nutrient with which it is paired (Yeomans et al. 2008). When novel flavors are paired with foods with higher energy content, learning more readily occurs, perhaps because it is adaptive to consume flavors that reliably predict energy content of foods. Foods that are higher in energy density tend to be higher in fat and sugar. Thus, flavor–nutrient learning may explain why high-fat/high-sugar foods are perceived as more palatable and are more highly liked (Drewnowski 1997).

Caffeine has been shown to increase liking of novel-flavored beverages. However, it is not known if caffeine can also increase liking of foods with which it is paired. The purpose of the current experiment was to test the hypothesis that a caffeinated beverage paired with a novel-flavored yogurt will increase preference for that yogurt relative to one paired with placebo. In addition, using the flavor–nutrient learning model as a theoretical framework, we tested the hypothesis that increased liking would be the strongest when caffeine was paired with high energy density yogurt. The outcome of this study could serve as a model for increasing liking and preferences for foods that are consumed with caffeinated beverages.

Methods

Participants

From a database of volunteers, 18- to 50-year-old women and men were recruited who had expressed interest in participating in future studies as well as through flyers distributed throughout the University at Buffalo campus and surrounding community. Exclusion criteria included medications or medical conditions that affected eating or were contraindicated (i.e., known pregnancy, allergies, etc.) and prior adverse reactions to caffeine. Eligible individuals were scheduled for six visits to the laboratory, occurring between 11:00 am and 5:00 pm, on separate days, with the second through fifth visits taking place on consecutive days. Each visit was approximately 1 h in duration. Upon scheduling, the participants were randomly assigned to one of two caffeine conditions (placebo (PLA) or caffeine (CAF, 2 mg/kg body weight)) and to one of two energy density conditions (low energy density (LED, 0.5 kcal/g) or high energy density (HED, 1.5 kcal/g)), with gender and usual caffeine use used as stratification variables to assure a balanced sample across study conditions. Participants were asked to abstain from all forms of caffeine 24 h prior to each laboratory session and any food or drink besides water 2 h prior to each session.

Baseline assessment

All procedures described within this manuscript were approved by the University at Buffalo Social and Behavioral Sciences Institutional Review Board. Upon their first visit to the laboratory, participants read and signed informed consent forms. Participants were told that “the purpose of the study is to determine how substances commonly found in soft drinks affect food liking and preferences.” In an effort to prevent preconceptions about the effects of caffeine or nutritive or nonnutritive sweeteners from altering experimental results, participants were told that the beverage and yogurts they would be consuming “may have levels of one or more of the following substances manipulated: sugar, aspartame, Splenda®, caffeine, or artificial coloring.” Participants then completed a demographic questionnaire followed by a researcher-guided 24-h dietary recall to confirm caffeine abstinence. Participants were asked to provide a 2.5-mL saliva sample into a sterile tube at the start of each test session which they were told would be analyzed for caffeine. Although these samples were not analyzed, they provided added incentive to comply with 24-h caffeine abstinence (Yeomans et al. 2008). Participants then completed questionnaires on the computer using SurveyMonkey™: Caffeine Use, Yogurt Consumption, and Behavioral Checklist (each described below). Participants were presented with a 350-mL beverage (none of the beverages contained CAF on the first or last visit) and were instructed that they had 5 min to consume the entire beverage. After finishing the beverage, participants were presented with 50-g servings of seven different novel yogurt flavors presented in the same order to each participant and in the same order at visits 1 and 6 (described below). Participants randomized to the LED group received LED yogurt for the taste tests and those randomized to the HED group received HED yogurt on these visits. Participants were instructed to taste each flavor, eat as much or as little of each as they liked, and take a drink of water between each sample in order to “taste each flavor better.” Yogurt consumption (in grams) was determined by weighing the container, spoon, and yogurt before and after yogurt consumption. The participant was instructed to fill out a yogurt rating for each flavor (described below) and to rank the yogurts from least favorite to most favorite. The yogurt which ranked fourth became the target flavor for each individual. We chose the fourth ranked yogurt because other studies have used the fourth ranked yogurt out of seven when conducting similar studies (Richardson et al. 1996; Temple et al. 2012) and we wanted a yogurt that could increase in liking and ranking but that would not be so disliked that participants would be unwilling to consume it. After the taste test and ranking, the participant again completed the Behavioral Checklist. At the end of the visit, the participant’s height and weight were measured to determine caffeine dosage and anthropometrics. Finally, a reminder slip was provided with date, time, and contact information for the next visit as well as a reminder to abstain from caffeine for 24 h prior to the visit. As all second visits were scheduled for Mondays, reminder calls were also made to each participant the Friday before their second visit with not only a reminder of time and place but also a reminder to abstain from caffeine for 24 h and fast for 2 h prior to the visit.

Exposure visits

For the next four laboratory visits (visits 2–5), participants came to the laboratory, provided a saliva sample, and completed a 24-h dietary recall and Behavioral Checklist. They were again given 5 min to consume the entire 350-mL beverage containing PLA or CAF. After consumption of the beverage, they were presented with a 100-g serving of their target yogurt, told to consume the entire serving, and asked to complete a yogurt rating questionnaire. When designing this experiment, we considered the optimal temporal relationship between presentation of the caffeine or placebo and the yogurt. One option would have been to present the yogurt 30 min after the caffeine so that it coincided with a peak in plasma caffeine levels. We chose to present them together, however, because our previous study had the novel flavor directly paired with the caffeine (Temple et al. 2012) and because we were examining the potential combined effects of flavor–nutrient learning along with caffeine and the flavor–nutrient learning paradigms pair the flavors directly with the foods of varying energy densities (Capaldi and Privitera 2007; Yeomans et al. 2008). After consumption of the yogurt, participants were asked to wait 30 min during which time they were allowed to engage in quiet activities, such as reading or listening to music. Magazines were provided if participants did not bring any activities of their own. After the 30 min, they again completed the Behavioral Checklist and were given a reminder slip for their appointment the following day.

Postexposure assessment

The sixth visit was identical to the first except that the participants were administered with additional paper questionnaires at the end: the Three-Factor Eating Questionnaire (TFEQ) (Bond et al. 2001), the Binge Eating Scale (BES) (Celio et al. 2004), and the Questionnaire of Weight and Eating Patterns (Celio et al. 2004) which assessed dietary restraint, disinhibition, and ruled out potential eating disorders, respectively (each is described below). They then completed a researcher-guided structured debriefing (described below), the nature of the experiment was disclosed, and the participants were compensated for their participation.

Yogurt preparations

The base HED yogurt (energy density = 1.5 kcal/g) was made by combining 900 g of FAGE® Total full-fat plain Greek yogurt (1,019 kcal, 36 g fat, 102 g carbohydrate, 96 g sugars, 66 g protein) with 130 g sucrose. The base LED yogurt (energy density = 0.5 kcal/g) was made by combining 900 g of Wegmans® 0 % fat plain Greek yogurt (478 kcal, 0 g fat, 40 g carbohydrate, 40 g sugars, 80 g protein) with 60 grams of Splenda®. To make the following seven novel combinations (raspberry–lemon, maple, almond, peppermint, pumpkin pie, strawberry–coconut, and cumin–turmeric), 50- (baseline and postexposure) or 100-g (exposure) portions were measured and flavored with McCormick® extracts and spices. Just prior to consumption, sample(s) were opened, stirred, and presented in matching 236-mL Ball® plastic containers and stainless steel teaspoons.

Beverage preparations

The base beverage, 350 mL Sprite®, was used without treatment at the baseline and postexposure visits as well as with treatment (PLA 0 mg/kg or CAF 2 mg/kg) for exposure visits. The treatment was prepared and coded by an investigator who was not otherwise involved with the study, and all research assistants on the study and participants were blind to the treatment. For placebo, Sprite® was heated to 60 °C and stirred at 50 rpm for 25 min, then pipetted into 10-mL aliquots into 14-mL polystyrene tubes and frozen at −20 °C. For CAF, powder anhydrous caffeine was added in the ratio of 10 mg/mL Sprite® and then heated, stirred, pipetted, and frozen the same as the placebo preparation. For the exposure visits, 2 mg/kg body weight of PLA or CAF was added to the base beverage.

Measurements

Anthropometrics

At the end of the baseline visit, weight was assessed by use of a digital scale (SECA; Hanover, MD). Height was assessed using a digital, wall-mounted stadiometer (SECA; Hanover, MD). On the basis of the height and weight data, BMI was calculated according to the following formula: (BMI, in kilogram per square meter). Body weight (in kilogram) was used to determine accurate dosage for exposure visits.

Twenty-four-hour dietary recall

To ensure compliance with the protocol (no food or drink other than water for 2 h prior to testing and no caffeine for 24 h prior to testing), at each visit, the participant completed a guided recall of his/her dietary intake for the previous day. In addition, the participant was asked if they had consumed any caffeine the current day. The experimenter guided the participant through the recall process using a five-step, multi-pass interview style. Briefly, the first pass included asking the participant to make a quick, uninterrupted list of all foods and beverages consumed. The second pass was a review of the quick list. For the third pass, the experimenter returned to the beginning of the list and asked for times that foods and beverages were consumed and for portion sizes. The fourth pass was a review of detailed information. During the final pass, the participant was asked to recall any other foods that they may have forgotten, such as foods that were eaten in small amounts. Measuring cups and spoons and rulers were provided to help the participants estimate portion sizes. Energy and macronutrient intakes were calculated using Nutritionist Pro nutrient analysis software.

Surveys/questionnaires

Caffeine use

This computer-based questionnaire assessed sources, amounts, and frequency of usual caffeinated beverage intake in our study population. It also assessed reasons why adults use and/or do not use caffeine. Amounts of caffeine consumed were estimated based on the information from the US Department of Nutritional Services and include the following: tea (40 mg/5 oz), soda (40 mg/12 oz), coffee (100 mg/5 oz), energy drinks (150 mg/12 oz), chocolate (10 mg/oz), and caffeine-containing pills (Excedrin or NoDoz—130–200 mg/pill).

Behavioral checklist

A computer-based questionnaire containing 31 adjectives (nervous, lively, happy with the way things are, sad, dizzy, falling asleep, my body feels tired, need to pee, headache, can’t sleep, irregular heartbeat, diarrhea, impatient, hungry, cranky, motivated to work, life is good, mood swings, muscle twitches, talkative, queasy, heart racing, can’t sit still, ringing in ears, energetic, stomach ache, strong, sweating, shaky, sleepy, tired) describing mood and physiological symptoms was presented to the subjects at the beginning and upon completion of each session. The subject was asked to rate how each adjective describes how they felt “right now” on a nine-point Likert-type scale anchored by “Not at all” (1) and “Extremely” (9). This questionnaire was adapted from the Profile of Mood States—Bipolar Form (Boyle 1987, 1988), the Activation–Deactivation Adjective Checklist (Guardia and Adan 1997), and DSM-IV symptoms of caffeine withdrawal (Hughes et al. 1992, 1998).

Yogurt rating and ranking

Participants were asked to provide subjective ratings of different qualities of the study yogurts. They were asked to draw a single vertical line indicating “how you feel right now” or “how much do you like the yogurt” as well as “how novel/pleasant/acidic/bitter/sour/spicy/sweet/strong is the yogurt flavor.” These were rated on a 100-mm visual analog scale (VAS) anchored by “Not at all” (0 mm) and “Extremely” (100 mm). In addition, subjects were asked to rank the yogurts from “Least Liked” (1) to “Most Liked” (7).

Analytical plan

Differences in participant characteristics between energy density (LED or HED) and treatment (PLA or CAF) groups were analyzed using analysis of variance for continuous data and chi-square test for categorical data with treatment group and energy density group as the between-subjects variables. Group differences in target yogurt ratings, rankings, and yogurt intake (in grams) were analyzed using a mixed analysis of covariance (ANCOVA) with energy density group (LED or HED) and treatment group (PLA or CAF) as between-subjects factors, trials as the repeated measure, and daily caffeine intake, baseline yogurt liking, and time of day that the participant completed the study as covariates. In order to examine potential changes in withdrawal symptoms across the duration of the study, we conducted a mixed ANCOVA with treatment group as between-subjects factors, trials as the repeated measure, and daily caffeine intake and time of day as covariates with pretreatment Behavioral Checklist responses on the following terms: dizzy, asleep, body tired, headache, mood swings, energy, shaky, sleepy, and tired. Because the Behavioral Checklist is meant to assess changes in mood and physiological state in response to caffeine, data were averaged across trials 2–5. Group differences in Behavioral Checklist responses were analyzed using a mixed ANCOVA with energy density group and treatment group as between-subjects factors, pre/posttreatment as the repeated measure, and daily caffeine intake and time of day as covariates. Significant interactions in the primary ANCOVA were probed by conducting separate two-way ANCOVAs to determine the source of the significance. All data were analyzed using SYSTAT 11.0 and data were considered significant if p < 0.05.

Results

Participants

Sixty-eight began and 62 completed the study; one participant who did not complete the study reported gastrointestinal upset, and the other five did not return due to scheduling conflicts or loss to follow-up. Five participants had visits rescheduled due to self-reported caffeine use on the study day. There were no significant differences between energy density or treatment groups for age, TFEQ, BES, daily caffeine intake, gender, marital status, education, employment status, student status, income, or ethnicity (all p > 0.092, Table 1). The distribution of daily caffeine use within the population was two-tailed, with 29 % considered very low to nonconsumers (<30 mg/day) and 27 % considered high users (>200 mg/day) with the remaining 44 % between 30 and 200 mg/day. There were no differences in the number of participants in these caffeine use categories as a function of treatment group or energy density group (all p > 0.2).
Table 1

Participant baseline characteristics

 

Placebo

Caffeine

p for treatment

p for energy density

HED

LED

HED

LED

N (% male)

12 (42 %)

13 (46 %)

14 (57 %)

11 (36 %)

0.77

0.98

 

Mean ± SEM

Mean ± SEM

Mean ± SEM

Mean ± SEM

  

Age

27.3 ± 2.1

31.1 ± 3.0

24.3 ± 1.6

29.3 ± 3.1

0.97

0.14

Restraint

10.8 ± 1.6

7.7 ± 1.3

9.2 ± 1.4

11.2 ± 1.2

0.64

0.36

Disinhibition

6.6 ± 1.4

5.2 ± 0.7

4.9 ± 0.9

5.0 ± 0.9

0.70

0.83

Hunger

4.4 ± 1.0

5.7 ± 0.8

4.7 ± 0.8

5.1 ± 1.2

0.79

0.24

BES

7.2 ± 1.6

8.6 ± 1.2

6.1 ± 1.5

8.8 ± 2.2

0.83

0.10

Mean of daily caffeine intake (mg)

157.5 ± 43.0

94.8 ± 26.8

120.2 ± 34.8

186.6 ± 36.7

0.49

0.91

Range of daily caffeine intake (mg)

3.2–547.5

0–274.04

1.8–568.6

15.5–451.3

N/A

N/A

 

N (%)

N (%)

N (%)

N (%)

  

Education

    

0.47

0.51

 Completed high School

1 (8)

0 (0)

0 (0)

1 (9)

  

 Some college/vocational

0 (0)

7 (54)

6 (43)

5 (45)

  

 Completed college/graduate degree

11 (92)

6 (46)

8 (57)

5 (45)

  

Income

    

0.75

0.61

 <$10,000/year

2 (17)

3 (23)

4 (29)

3 (27)

  

 $10,000–$99.999/year

7 (58)

6 (46)

5 (36)

6 (55)

  

 >$100,000/year

1 (8)

1 (8)

0 (0)

1 (9)

  

 No response

2 (17)

2 (15)

5 (36)

1 (9)

  

Ethnicity

    

0.56

0.78

 White/Caucasian

9 (75)

5 (38)

5 (36)

9 (82)

  

 Asian

2 (17)

4 (31)

7 (50)

1 (9)

  

 Black/African-American

1 (8)

3 (23)

1 (7)

0 (0)

  

 No response

0 (0)

1 (8)

1 (7)

1 (9)

  

Caffeine and energy density affect yogurt liking

Baseline yogurt liking differed as a function of energy density, with HED yogurt liked significantly more than LED yogurt (F(1, 58) = 4.5, p = 0.04). Because of this, we used baseline yogurt liking as a covariate in all analyses. The change in liking of yogurt was analyzed over the 6 days of exposure, controlling for differences in baseline yogurt liking. We found main effects of trials (F(5, 240) = 30.9, p < 0.0001) on yogurt liking, with yogurt liking increasing on subsequent trials in all groups. There was a two-way interaction between trials and caffeine treatment (F(5, 240) = 2.8, p = 0.017), with liking of yogurt increasing over time more when participants were given caffeine as compared to placebo. There was a three-way interaction among trials, yogurt energy density, and caffeine treatment (F(5, 240) = 3.3, p = 0.007). Separate two-way ANCOVAs on LED and HED yogurt conditions revealed that CAF treatment had no effect in the individuals who consumed the HED yogurt (F(5, 125) = 0.92, p = 0.47), but significantly increased yogurt liking in the individuals consuming LED yogurt (F(5, 115) = 3.3, p = 0.008; Fig. 1). None of these effects were influenced by usual caffeine intake or time of day.
https://static-content.springer.com/image/art%3A10.1007%2Fs00213-013-2971-6/MediaObjects/213_2013_2971_Fig1_HTML.gif
Fig. 1

Mean ± SEM liking ratings for a novel-flavored yogurt in participants who had a high energy density (HED) yogurt or b low energy density (LED) yogurt. The values shown are adjusted for differences in baseline liking because we found that participants rated the HED yogurt more highly liked than the LED yogurt. Repeated measures ANCOVA revealed that the pairing of the HED yogurt with caffeine had no effect on yogurt liking, but significantly increased liking of LED yogurt relative to pairing with placebo (p = 0.004)

Caffeine treatment and energy density affect yogurt intake

Amount (in grams) of the target yogurt consumed was analyzed at baseline and follow-up. We found a main effect of pre/posttreatment (F(1, 56) = 13.6, p = 0.001) and yogurt energy density (F(1, 56) = 4.2, p = 0.044) on grams consumed, with grams consumed increasing from baseline to postexposure and with more grams of HED yogurt consumed than LED yogurt. There were two-way interactions between pre/posttreatment and yogurt energy density (F(1, 56) = 8.5, p = 0.005) and pre/post and caffeine treatment (F(1, 56) = 5.6, p = 0.02). Separate ANCOVA analyses revealed that grams of yogurt consumed increased more for the HED than LED yogurt (F(1, 58) = 7.5, p = 0.008; Fig. 2b) and grams consumed increased more in the CAF group than in the PLA group (F(1, 58) = 6.2, p = 0.016; Fig. 2a) from baseline to postexposure. We also found that yogurt consumption increased for some of the nonexposed yogurts. There was a main effect of pre/posttreatment on consumption of raspberry–lemon (F(1, 43) = 12.4, p = 0.01), maple (F(1, 49) = 7.6, p = 0.008), and strawberry coconut (F(1, 51) = 7.7, p = 0.008), but not any of the other flavors. There were no effects of yogurt energy density or caffeine treatment on changes in ranking of yogurts from visit 1 to visit 6 (all p > 0.45, Table 2). Daily caffeine intake and time of day had no effect on these findings.
https://static-content.springer.com/image/art%3A10.1007%2Fs00213-013-2971-6/MediaObjects/213_2013_2971_Fig2_HTML.gif
Fig. 2

Mean ± SEM consumption of yogurt (in grams) at baseline and again after the exposure trials (6–7 days after baseline) in participants who consumed yogurt paired with either placebo or caffeine (a) and in participants who consumed yogurt that was low energy density (LED) or high energy density (HED) (b). ANCOVA analyses revealed that there was a main effect of caffeine exposure on consumption of yogurt, with more consumption of the yogurt paired with caffeine compared with the one that was paired with placebo (p = 0.003). There was also a main effect of yogurt energy density on consumption of yogurt, with participants in the HED group consuming more yogurt than those in the LED group (p = 0.016). There were no interactions between caffeine condition and energy density condition for consumption measures

Table 2

Rating of taste properties and rankings of different yogurt flavors

 

Like

Novel

Pleasant

Acidic

Bitter

Sour

Spicy

Sweet

Strong

Rank

1

6

1

6

1

6

1

6

1

6

1

6

1

6

1

6

1

6

1

6

Raspberry–lemon

54.7 (3.6)

64.4 (2.6)

51.8 (3.6)

50.7 (3.6)

54.9 (3.5)

65.2 (2.6)

39.2 (3.4)

25.3 (3.1)

33.2 (3.4)

22.1 (2.6)

35.4 (3.5)

30.7 (3.3)

6.2 (1.1)

8.6 (1.8)

62.7 (2.6)

58.1 (2.5)

58.2 (3.3)

59.8 (2.6)

5.1 (0.2)

4.9 (0.2)

Maple

62.5 (3.7)

74.5 (2.2)

53.2 (3.3)

54.5 (3.1)

60.7 (3.4)

74.3 (2.2)

20.2 (2.6)

14.8 (2.1)

23.3 (2.9)

14.5 (2.1)

22.6 (3.2)

16.5 (2.5)

5.9 (0.9)

8.3 (1.7)

61.9 (2.9)

58.6 (2.3)

61.2 (2.7)

62.7 (2.)

5.4 (0.2)

5.9 (0.2)

Almond

36.1 (3.7)

41.8 (3.5)

58.4 (3.6)

60.5 (3.5)

38.2 (3.7)

44.2 (3.8)

27.8 (3.2)

27.6 (3.5)

34.5 (3.9)

26.8 (3.5)

36.1 (3.7)

29.3 (3.5)

7.2 (1.3)

11.2 (2.1)

53.1 (3.0)

51.4 (2.8)

71.8 (2.6)

70.3 (2.7)

3.7 (0.2)

3.8 (0.2)

Peppermint

18.1 (3.1)

19.6 (2.8)

68.6 (4.0)

70.0 (3.2)

19.3 (3.0)

23.4 (3.5)

34.3 (4.2)

32.3 (3.9)

40.7 (4.4)

32.3 (3.9)

27.9 (3.7)

26.6 (3.3)

21.1 (3.8)

22.6 (3.8)

37.5 (3.3)

36.1 (3.2)

86.3 (2.0)

86.7 (1.8)

2.3 (0.2)

2.3 (0.2)

Pumpkin pie

44.9 (3.7)

45.0 (3.5)

67.2 (3.5)

64.9 (3.2)

46.2 (3.5)

45.9 (3.5)

25.5 (3.2)

21.1 (3.1)

28.9 (3.7)

24.8 (3.3)

24.9 (3.3)

20.4 (2.8)

32.4 (3.9)

35.1 (3.8)

47.5 (3.1)

46.6 (3.0)

72.3 (2.5)

73.6 (2.5)

4.0 (0.2)

3.9 (0.2)

Strawberry–coconut

66.3 (2.9)

68.3 (2.8)

40.2 (3.4)

42.4 (3.7)

69.0 (2.6)

71.2 (2.3)

20.2 (2.3)

14.9 (1.9)

14.8 (2.0)

13.7 (2.2)

23.8 (3.1)

17.9 (2.6)

6.7 (1.3)

6.1 (1.0)

67.6 (2.5)

64.9 (2.5)

58.7 (3.2)

60.7 (2.6)

5.7 (0.2)

5.5 (0.2)

Cumin–tumeric

17.5 (2.8)

14.4 (2.7)

69.6 (4.1)

75.4 (3.7)

17.8 (2.8)

15.5 (2.9)

29.8 (3.5)

28.2 (3.7)

49.7 (4.3)

40.2 (4.1)

33.5 (4.0)

27.5 (3.7)

56.3 (3.6)

58.4 (4.2)

20.6 (2.9)

20.1 (2.8)

79.2 (2.4)

83.5 (2.4)

1.9 (0.2)

1.8 (0.2)

Caffeine energy density, and yogurt taste properties

In addition to yogurt liking, we had the participants rate several taste properties of the target yogurt on each visit. We found a main effect of trials on ratings of novelty (F(5, 240) = 5.0, p < 0.0001), pleasantness (F(5, 240) = 26.8, p < 0.0001), acidity (F(2, 240) = 3.5, p = 0.005), bitterness (F(5, 240) = 4.2, p = 0.001), sourness (F(5, 225) = 2.5, p = 0.033), sweetness (F(5, 225) = 9.9; p < 0.0001), and strongness (F (5, 225) = 8.3, p < 0.0001). We also found two-way interactions between trials and caffeine treatment on pleasantness (F(5, 240) = 2.7, p = 0.02), sourness (F(5, 225) = 2.8, p = 0.017), and spiciness (F(5, 225) = 2.4, p = 0.04). Finally, we found a three-way interaction among trials, yogurt energy density, and caffeine treatment on yogurt acidity (F(5, 240) = 3.3, p = 0.007). Daily caffeine intake and time of day had no effect on these findings (Table 3).
Table 3

Yogurt taste ratings across trials as a function of energy density and treatment groups

Yogurt group

Treatment

 

Visit 1

Visit 2

Visit 3

Visit 4

Visit 5

Visit 6

LED

Placebo

Novela

49.8 ± 7.7

50.8 ± 7.2

40.3 ± 7.5

39.7 ± 6.3

39.5 ± 6.5

42.7 ± 7.3

HED

60.0 ± 8.7

60.6 ± 8.2

47.5 ± 9.2

50.9 ± 10.6

48.1 ± 9.4

47.0 ± 9.3

LED

Caffeine

45.7 ± 6.8

51.4 ± 5.9

53.9 ± 6.0

45.3 ± 6.9

52.4 ± 7.2

48.9 ± 7.5

HED

71.9 ± 4.7

65.1 ± 5.2

62.5 ± 5.6

57.8 ± 5.8

57.4 ± 5.9

63.5 ± 5.6

LED

Placebo

Pleasanta, b

35.7 ± 7.3

51.1 ± 6.6

49.5 ± 6.4

52.5 ± 6.6

51.1 ± 6.8

59.9 ± 5.8

HED

55.9 ± 6.3

77.1 ± 4.2

68.3 ± 6.3

77.3 ± 4.8

77.0 ± 5.3

74.1 ± 4.2

LED

Caffeine

36.6 ± 5.7

44.8 ± 6.9

58.0 ± 6.3

58.2 ± 6.8

68.5 ± 5.2

63.3 ± 6.8

HED

48.5 ± 6.0

57.0 ± 6.8

67.1 ± 5.2

68.4 ± 5.2

68.6 ± 4.0

71.8 ± 5.7

LED

Placebo

Acidica, c

32.4 ± 7.1

19.4 ± 6.3

24.5 ± 6.7

13.7 ± 4.6

21.7 ± 6.6

18.3 ± 5.5

HED

21.4 ± 5.4

28.0 ± 7.1

28.7 ± 7.0

17.5 ± 4.5

13.4 ± 2.9

15.9 ± 4.6

LED

Caffeine

49.5 ± 7.1

32.0 ± 6.7

34.6 ± 6.8

32.9 ± 6.8

23.4 ± 4.7

23.9 ± 7.0

HED

35.0 ± 6.3

25.4 ± 5.2

22.6 ± 5.0

20.9 ± 4.9

28.1 ± 5.1

25.5 ± 6.3

LED

Placebo

Bittera

33.3 ± 7.7

38.2 ± 7.6

30.4 ± 6.3

27.5 ± 6.5

29.9 ± 6.8

22.3 ± 5.6

HED

19.1 ± 5.8

17.0 ± 6.1

16.1 ± 5.3

9.2 ± 2.9

10.6 ± 3.0

9.4 ± 4.4

LED

Caffeine

44.1 ± 7.4

42.5 ± 6.8

38.2 ± 6.9

34.7 ± 7.2

31.4 ± 6.2

32.8 ± 7.2

HED

30.8 ± 5.4

30.5 ± 6.2

23.8 ± 5.4

22.1 ± 5.6

21.7 ± 4.7

18.3 ± 4.5

LED

Placebo

Soura, b

32.4 ± 8.2

32.3 ± 7.9

29.5 ± 6.0

20.4 ± 4.8

24.4 ± 5.5

22.1 ± 6.9

HED

22.4 ± 5.9

19.6 ± 5.9

25.8 ± 7.7

21.2 ± 5.9

20.2 ± 7.1

13.2 ± 5.1

LED

Caffeine

52.8 ± 7.7

30.8 ± 7.6

33.8 ± 6.7

33.1 ± 7.2

26.8 ± 6.1

36.7 ± 7.0

HED

31.3 + 6.4

27.5 ± 5.1

26.5 ± 5.8

27.6 ± 5.8

25.4 ± 4.7

29.3 ± 6.4

LED

Placebo

Spicyb

16.3 ± 5.9

8.5 ± 3.1

7.5 ± 4.0

7.3 ± 3.9

9.9 ± 5.8

12.5 ± 5.1

HED

5.1 ± 1.5

5.8 ± 1.9

3.9 ± 1.0

3.2 ± 0.8

2.3 ± 0.5

3.1 ± 0.6

LED

Caffeine

22.6 ± 7.7

18.6 ± 6.2

18.8 ± 6.4

22.1 ± 8.1

14.0 ± 6.0

25.0 ± 7.9

HED

10.2 ± 3.4

20.1 ± 5.4

12.9 ± 4.0

14.8 ± 4.5

18.3 ± 4.4

17.9 ± 5.2

LED

Placebo

Sweeta

48.3 ± 6.4

52.2 ± 6.0

51.5 ± 5.9

54.3 ± 5.6

58.1 ± 5.2

48.8 ± 5.9

HED

58.9 ± 6.4

64.4 ± 6.2

59.6 ± 6.2

61.4 ± 6.6

57.8 ± 6.3

63.7 ± 4.7

LED

Caffeine

51.8 ± 7.9

58.1 ± 5.2

64.2 ± 4.2

57.2 ± 5.0

54.8 ± 6.0

55.5 ± 5.5

HED

59.1 ± 5.6

57.1 ± 4.8

57.8 ± 5.0

64.6 ± 4.0

55.8 ± 4.4

62.8 ± 4.2

LED

Placebo

Stronga

58.9 ± 6.4

50.4 ± 7.0

50.1 ± 6.3

48.7 ± 5.5

49.9 ± 5.9

59.1 ± 5.1

HED

66.2 ± 6.4

68.8 ± 5.9

66.2 ± 4.9

64.5 ± 5.5

68.2 ± 4.9

67.1 ± 5.7

LED

Caffeine

71.2 ± 6.7

71.1 ± 3.9

72.1 ± 3.9

67.5 ± 4.3

59.8 ± 4.2

72.3 ± 2.6

HED

65.3 ± 5.9

62.0 ± 5.1

54.0 ± 6.3

59.9 ± 6.4

58.6 ± 5.3

69.8 ± 3.9

aSignificant change over trials

bSignificant interaction of caffeine treatment and trials

cSignificant interaction among trials, caffeine treatment, and yogurt energy density

Time and caffeine affect behavioral checklist measures

Changes in mood state were averaged pre- and posttreatment over the four exposure days. There were main effects of pre/posttreatment on feeling sad (F(1, 60) = 8.4, p = 0.005), feeling hungry (F(1, 60) = 55.7, p < 0.0001), feeling talkative (F(1, 60) = 9.0, p = 0.004), feeling unable to sit still (F(1, 60) = 6.5, p = 0.013, feeling strong (F(1, 60) = 4.9, p = 0.03), and feeling sweaty (F(1, 60) = 7.0, p = 0.01) with feeling sad, hungry, talkative, strong, and sweaty decreasing and feeling unable to sit still increasing from pre- to posttreatment. None of these responses were influenced by CAF treatment or usual CAF intake. There were also two-way interactions between pre/post and CAF treatment on feeling like falling asleep (F(1, 60) = 12.0, p = 0.001), body feeling tired (F(1, 60) = 5.4, p = 0.023), feeling heart racing (F(1, 60) = 5.7, p = 0.02), feeling shaky (F(1, 60) = 4.2, p = 0.046), and feeling sleepy (F(1, 60) = 6.3, p = 0.015) with feeling like falling asleep, body feeling tired, and feeling sleepy increasing with PLA treatment and decreased with CAF treatment and feeling heart racing and feeling shaky increasing in CAF group with no change in the PLA group from pre- to posttreatment (Table 4). When we analyzed baseline Behavioral Checklist responses across trials to determine if participants experienced withdrawal throughout the duration of the experiment, we found no differences as a function of treatment group or time in any of the responses examined (all p > 0.15).
Table 4

Results from the Behavioral Checklist

 

Placebo

Caffeine

p value

Pre

Post

Pre

Post

Mean ± SEM

Mean ± SEM

Mean ± SEM

Mean ± SEM

Asleepa

1.93 ± 0.2

2.67 ± 0.3

2.09 ± 0.2

1.96 ± 0.2

0.001

Body feels tireda

2.52 ± 0.3

2.65 ± 0.3

2.76 ± 0.2

2.37 ± 0.2

0.023

Heart racinga

1.18 ± 0.1

1.12 ± 0.1

1.18 ± 0.1

1.48 ± 0.2

0.02

Hungryb

4.78 ± 0.3

2.99 ± 0.3

4.52 ± 0.3

3.15 ± 0.2

<0.0001

Sadb

1.38 ± 0.1

1.22 ± 0.1

1.90 ± 0.2

1.65 ± 0.2

0.005

Shakya

1.26 ± 0.1

1.24 ± 0.1

1.22 ± 0.1

1.65 ± 0.2

0.046

Sleepya

2.37 ± 0.2

3.04 ± 0.3

2.63 ± 0.3

2.55 ± 0.2

0.015

Strongb

4.70 ± 0.3

4.57 ± 0.3

4.77 ± 0.3

4.60 ± 0.3

0.03

Sweatyb

1.47 ± 0.2

1.23 ± 0.2

1.46 ± 0.2

1.27 ± 0.1

0.01

Talkativeb

3.89 ± 0.3

3.36 ± 0.3

3.26 ± 0.3

3.10 ± 0.3

0.004

Unable to sitb

1.41 ± 0.1

1.59 ± 0.2

1.60 ± 0.2

1.88 ± 0.2

0.013

Mean ± SEM ratings of each adjective before and after placebo or caffeine treatment on a nine-point scale with 1 being “Not at all” and 9 being “Extremely”

aSignificant interaction between caffeine treatment and pre/posttreatment

bSignificant effect of pre/posttreatment

Discussion

The purpose of the current experiment was to test the hypothesis that caffeine paired with novel-flavored yogurt would increase liking of that yogurt more than when it was paired with placebo. Caffeine treatment increased yogurt liking regardless of usual caffeine intake, suggesting that this shift in liking may not rely on withdrawal reversal, as has been shown in previous studies (Chambers et al. 2007; Yeomans et al. 1998, 2002b, 2005). We also hypothesized that the increased liking would be strongest when caffeine was paired with HED yogurt. Contrary to our hypothesis, we saw liking of HED yogurt increased over time, regardless of caffeine treatment, but caffeine increased liking of LED yogurt more than placebo-paired LED yogurt. These effects were also not influenced by usual caffeine intake. Taken together, these findings show that caffeine administration may work independently of usual caffeine consumption and may have more influence over liking of foods that are not highly liked at baseline.

Based on our previous work from our laboratory and others, we expected to find an increase in liking of novel-flavored yogurt with caffeine treatment. Yogurt liking increased over time, and this increase was related to both energy density and treatment. The primary finding of the study was participants consuming LED yogurt paired with caffeine had a significant increase in liking for the novel yogurt flavor relative to those with LED yogurt paired with placebo. To our knowledge, no other studies have examined the effects of caffeine on liking when paired with a food. One previous study showed liking of a colored caffeinated beverage generalized to a color-matched sweet, fat candy (M&M®), which was indicated by a preference for that color of candy; however, neither liking, pleasantness, nor intake for the candy was evaluated (Dack and Reed 2009). Based on previous works demonstrating that liking of novel flavors increases after repeated exposure and pairing with HED foods (O’Sullivan et al. 2010; Yeomans et al. 2008; Appleton et al. 2006), we predicted that the HED yogurt liking would increase with repeated exposure but that the combination of caffeine and HED would increase the liking further. Instead, we saw a main effect of energy density on liking, with flavors paired with the HED yogurt increasing in liking, which is consistent with flavor–nutrient learning, with no further enhancement by caffeine treatment. The most likely explanation for this is that there was a ceiling effect and the HED yogurt was unable to be liked more than what was induced by energy density; however, we cannot rule out other potential explanations. The results from the placebo group are consistent with previous studies on flavor–nutrient learning, with little change in yogurt liking in the LED yogurt with the placebo. The caffeine pairing with the LED yogurt did increase liking over time in a manner that was not distinguishable from the HED yogurt groups. This suggests that caffeine can increase liking of foods with which it is paired. It is useful to note here that these results suggest that caffeine and energy density work in parallel to influence flavor preferences and that energy density may play a more important role than caffeine. It would be interesting to know if the role of caffeine would be strengthened if the context in which it was paired was one in which participants would typically expect to receive caffeine. Future studies need to determine the role of expectancy in the effects of caffeine on flavor preferences.

We also found that intake of the target yogurt increased from the first to the last visit. The magnitude of increase was related to both energy density and caffeine treatment independently, with no interactions between the two. This is consistent with other studies that have shown that repeated exposure to a novel food or flavor of food can increase acceptance of that food (Anzman-Frasca et al. 2012; Wardle et al. 2003). We were not surprised that the HED yogurt consumption increased more than the LED yogurt consumption, as it was reported to be liked more at baseline. When we examined changes in ranking of the target yogurt, we did not find any differences as a function of energy density or caffeine treatment. This suggests that although liking increased and consumption increased, when the flavor was presented among other flavors that were originally liked more and less, the ranking among the seven yogurts remained the same. We have reported similar findings in a previous study examining the effects of caffeine on liking of novel-flavored soda (Temple et al. 2012). There may be a difference in the perception of the food or beverage when it is presented on its own compared with when it is presented among other flavors. Another explanation is that the preexposure to caffeine influenced the flavor perception of the yogurt during the exposure visits. Because we did not control for the bitter taste in the caffeinated beverage, participants receiving caffeine experienced a slightly bitter tasting drink immediately prior to the yogurt exposure, while those receiving placebo did not experience this bitterness. We found that the caffeine group reported their yogurt to be more acidic, more sour, and spicier than those in the placebo group. Future studies will control for preexposure to bitterness when examining changes in taste preferences. It should be noted, however, that despite the higher ratings for unpleasant taste characteristics of the yogurt in the caffeine group, this group reported higher pleasantness and higher liking than the placebo group.

Previous studies have shown that the influence of caffeine on liking of novel-flavored beverages depends upon reversal of withdrawal symptoms resulting from caffeine abstinence (Richardson et al. 1996; Rogers et al. 1995; Yeomans et al. 1998, 2002b, 2005, 2007). In our study, the increase in liking of LED yogurt induced by caffeine administration occurred independent of usual caffeine consumption. There are several potential explanations for this. First, we found that liking of the flavors in the HED yogurt increased over time regardless of caffeine treatment. In previous studies (Yeomans et al. 1998, 2002b, 2005, 2007), the liking of the novel beverage flavors did not increase when paired with placebo, suggesting that our flavors may have been more liked to begin with. Second, our population had a broad range of caffeine consumption (0–570 mg/day), but the average was lower than what would be expected in an adult population (135 ± 16.7 mg/day), which suggests that our population, on average, may have been low to moderate users of caffeine, which reduces the likelihood of caffeine withdrawal. Third, because our participants were required to visit the laboratory for four consecutive days for exposure and were required to be abstinent from caffeine continuously for 4–5 days, this could have affected our findings in a number of ways. For example, even light consumers of caffeine may have been experiencing some withdrawal symptoms after this many days of abstinence. Conversely, by the end of the study, participants may have no longer been experiencing any withdrawal symptoms, which may explain why the group differences are diminished by the final day of the study. In order to eliminate the possible influence of withdrawal from caffeine, future studies may need to employ a long-term withdrawal period prior to the beginning of the study. Finally, it is possible that the flavors of our yogurts were not as novel as flavors used in previous studies. Although the flavors that we used are not found in yogurts, they are flavors that people are familiar with (peppermint, pumpkin pie spice, etc.). Our average novelty ratings at baseline were 58.2 ± 3.3 on a 100-mm VAS. This is lower than what has been reported in other studies. This prior experience with the flavors in a different context may have made them more readily liked. We do not think that this is the most plausible relationship, as prior studies have shown that high levels of novelty are important to achieve this increase in liking (Smit et al. 2006). Therefore, we would have expected that familiarity with the flavors would have reduced the ability of energy density and/or caffeine to alter liking. These results suggest that pairing of caffeine with moderately novel and moderately liked foods may increase liking of those foods, regardless of prior caffeine consumption or withdrawal.

This study had several strengths including a large sample size (n = 62) with no significant differences in characteristics between groups (p > 0.05). In addition, the study was a double-blind randomized controlled trial with the researchers and subjects blinded to treatment (PLA/CAF) and subjects blinded to yogurt group (HED/LED). This study was not without limitations. First, the baseline novelty (58.2 ± 3.3) of the yogurt flavors was lower than what other studies have reported and may not have been optimal for maximizing changes in liking (Yeomans et al. 2005). Second, because of the experimental design, it is not possible to attribute our findings to conditioning, as previous studies have been able to do so. In order to achieve this, we would have had to conduct a within-subjects design where one flavor of yogurt was paired with caffeine and a different flavor of yogurt was paired with placebo, as has been done in previous studies (Yeomans et al. 2005). Future studies will utilize this design so that we may test the hypothesis that caffeine conditions taste preferences to food with which it is paired. Third, it is possible that because participants were instructed to refrain from caffeine-containing products for the duration of the study that they might have been aware that caffeine was being manipulated. It is well known that expectancy about caffeine’s effects can influence study outcomes (Huntley and Juliano 2012). Fourth, because both placebo and caffeine-containing beverages contained nutritive sweeteners, we cannot rule out the possibility that the influence of the beverage on yogurt liking was related, at least in part, to the increase in sugar. We do not believe that this explains our results entirely, as the placebo beverage was sweetened with sucrose as well, but it is possible that the increase in liking of the LED yogurt was related to synergistic effects of caffeine and sucrose. Fifth, although we collected saliva from participants and told them that it would be used for analysis of caffeine to confirm abstinence, we did not do this. Therefore, we have no biochemical verification of caffeine abstinence. Finally, we did not use a bitter flavor in the placebo; thus, participants may have had differences in liking of the beverage, as the caffeinated beverage may have been perceived as more bitter than the placebo beverage. The bitterness in the beverage may have influenced the ratings of the yogurt that followed it. In addition, because participants were told that the caffeine could have been manipulated, the bitterness of the caffeinated beverage may have reduced the effectiveness of our strategies to maintain double-blind experimental procedures. Future studies will use a bitter tastant, such as quinine, in the placebo in order to match the beverages for bitter sensation and perception.

In conclusion, this study showed that caffeine added to a beverage increased liking and pleasantness of LED yogurts with LED yogurt liking increasing to a level equivalent to HED liking. In addition, caffeine increased the amount of LED yogurt consumed to a level similar to the amount of HED food consumed. Future studies will focus on testing the impact of pairing caffeine with other LED foods to see if this increase in liking will generalize to a broader range of foods. In addition, future studies will focus on determining the optimal temporal pairing of caffeine with foods to achieve the greatest changes in liking. Finally, we will also test the potential synergy between caffeine and caloric sweeteners on liking of foods and beverages by using a nonnutritive sweetener comparison group. While there is still much work to be done in this area, we feel that these findings are promising and merit further study.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013